THD and all that, since Hugh made reference above.
Back in 1997 Lynn Olsen wrote an excellent article on amplifier distortion (Glass Audio Vol9 No 4)
In his introduction he wrote:
"....the subjective correlation between the total harmonic distortion (THD) measurement and what you actually hear is close to zero".
He then went on to give an excellent analogy:
"The fault is not with the subjective perception of the listener, but with the measurement itself. There is nothing very new in this; you can measure all you want, but a mass spectrometer is not going to find a lot of difference between lunch at a high school cafeteria and the best dinner at a four-star restaurant. To foolishly assert that the mass-spec. machine is right,.....,is an example of simple ignorance trying to cover its nakedness with a fig leaf of science."
The following summarises, in my own words, what he had to say.
To attempt to make some correlations between harmonic distortion measurements and perceived sound you need to first separate Even harmonics and Odd harmonics and view these as separate sets of data.
Even harmonics are generated by asymmetrical distortion mechanisms and odd harmonics are generated by symmetrical distortion mechanisms.
As an example, look at the differential amplifier on the front end of the AKSA 55, the LIFEFORCE and just about every SS amp from the last 30 years. To reduce the Even Harmonics you must balance the currents in the 2 transistors such that there is little or no asymmetry. Once this is done the residual distortion will be odd harmonics due to the symmetrical nonlinearities in the 2 devices.
This fundamental concept is important, and we get some big clues when we extend are consideration of distortion to include Intermodulation Distortion (IM).
Let us, for example, consider 2 frequencies in the important 1 to 5 kHz region.
For example use 3kHz and 4kHz:
Most of us understand that superimposing these (mixing in a non- linear system) will produce new frequencies and we often refer to these as sidebands. This is too simplistic.
The maths works like this:
Take 2 sinusoidal signal (well we use cosine rather than sine to keep the maths simple)
Signal 1 = a1(cos x)
Signal 2 = a2(cos y)
From output = a1 cos(x) + a2 cos(y)
Exapanding this and substituting in the trigonometric identity cos(x) + cos(y) = 1/2(cos(x+y) + cos(x-y)) and a lot of tedious algebra later we end up with an expression for:
1) the 2nd order term , For the superposition of 2 signals of x = 3KHz and y = 4kHz then four(4) new frequencies are created, 2 off 2nd harmonic terms and 2 off IM sidebands :
2x = 6kHz , 2nd harmonic of x
2y = 8kHz , 2nd harmonic of y
x + y = 7kHz , IM sideband
x - y = 1kHz , IM sideband
2) The 3rd order term gives six (6) new frequencies. 2 off 3rd harmonics and 4 off IM sidebands
3x= 9kHz , 3rd harmonic of x
3y = 12kHz , 3rd harmonic of y
2x + y = 10kHz , IM Sideband
2x - y = 2kHz , IM Sideband
2y + x = 11kHz , IM sideband
2y - x = 5kHz , IM Sideband
The important thing to note here is that the 2nd harmonic distortion results in new intermodulation product frequencies (sidebands) which are remote from the original frequencies. (1kHz and 7kHz)
The 3rd harmonic results in more intermodulation products, two(2) of which are very close to the original frequencies. (2kHz and 5kHz)
To exaggerate this (but still a valid example) see what happens with 14kHz and 15kHz
2nd order IM terms are 1kHz and 29kHz (very remote from the original 14 and 15kHz)
3rd order products are 13kHz and 16kHz (very close to the original 14 and 15 kHz)
Higher order terms produce even more IM Sidebands
I might be wrong (because I did'nt extend the algebra past the 2nd and 3rd order terms) but I think the 4th order will produce 6 IM sidebands and the 5th will produce 8 IM Sidebands.
Then consider other sources of a signal with which we can intermodulate.
One of the most important will be the 100Hz (120Hz in the US etc) residual power supply ripple which depending upon the quality of your power supply may well have 200, 300, 400Hz etc. harmonics as well.
At this point the "stray" IM products can run to thousands.
This is why a no feedback single ended triode amp with 2% of 2nd harmonic distortion but no 3rd, 4th, 5th etc can sound stunning and why an SS amp with 0.001% THD distortion which is primarily odd (and high) order distortion can sound awful.
It is also why some amplifiers which sound lovely with folk or jazz music which have a lower number of simultaneous tones (sparse spectra) can sound seriously rubbish when reproducing large choirs and orchestras (many closely spaced tones).
So what can we do about it?
The answer is simply to follow the established "popular wisdom".
1) Keep Power Supplies super clean
2) Don?t allow high order harmonic distortion products
3) What harmonic distortion there is should be even harmonics ONLY
So here is my theory / explanation / WAG,
In short, higher order harmonic distortions produce many more Intermodulation Sidebands
Even harmonic distortions create IM Sidebands remote from the original frequencies
Odd harmonic distortion produce IM Sidebands close to the original frequencies.
Any of this make sense to anyone?
Cheers,
Ian (Ginger)
